Introduction
Salmonella enterica serotype Heidelberg (S. Heidelberg) is a group B Salmonella which apparently accounts for a small proportion of cases of disease in humans. Between 1994 and 1997, S. Heidelberg was the tenth most frequently identified serotype isolated from humans in Italy (1.3% of all human isolates), although it does not appear among the top ten serotypes from the data from the Enter-Net Italia surveillance system available for 1999, 2000 and 2001. National data from the veterinary surveillance system of Salmonella serotypes do however confirm the presence of S. Heidelberg in the poultry farm environment, this serotype being identified in 2002 from 6.5% and 20.3%, respectively, of Salmonella strains from chicken and turkey (1).

Reports from countries including the United States (US) and Canada describe a high prevalence of the Heidelberg serotype in both human and non-human sources, mainly food and livestock (2-4). A nursing home outbreak attributable to this serotype, associated with Campylobacter jejuni, has also been reported by Layton et al (5).
In January 2002, a small family cluster of infections of S. Heidelberg, involving two children and their asymptomatic father, who was working in a poultry slaughterhouse, occurred in Pistoia, Tuscany. The strain exhibited plasmid-mediated resistances to ampicillin and sulphonamides. At the same time, in Palermo, Sicily, one S. Heidelberg strain with identical drug susceptibility pattern and plasmid profile was isolated from retail commercial chicken entrails. Between March and June 2003, S. Heidelberg was again identified at the Centro per gli Enterobatteri Patogeni per l'Italia Meridionale (Centre for enteric pathogens for southern Italy, CEPIM), from two paediatric cases and two samples of whole chicken on sale.
A typing study by phenotypic and molecular techniques was thus performed to substantiate a possible relationship between S. Heidelberg isolates from these two areas of Italy. A retrospective analysis of available isolates identified in the years 1999-2001 from different regions of Italy was also carried out.

Materials and methods
The study included 21 isolates from human cases, chicken products and the environment from different regions of Italy (Table). Epidemiological data about possible exposures to contaminated sources (eating of particular foods, previous travel or farm visit) were unavailable for most cases. All samples of chicken were on sale in Palermo and Agrigento (Sicily), but processed and marketed by two poultry factories with nationwide distribution channels.
Susceptibility patterns were assessed by disk diffusion according to the criteria of the National Committee for Clinical Laboratory Standards (6). Plasmid content was investigated by the alkaline lysis method (7). PFGE was performed by a standard procedure (8). The DNA size standard used was the bacteriophage lambda ladder ranging from 48.5 to 1.000 kb (Bio-Rad). Macrorestriction fragment patterns were visually
analyzed and classified by previously established guidelines (9, 10). Isolates with electrophoretic patterns differing by one to four DNA fragments were classified as subtypes of the same pulsetype. For the purpose of this study, different pulsotypes were considered to identify distinct clones of the Heidelberg serotype.

Results
Five distinctive antimicrobial resistance profiles were identified among the S. Heidelberg isolates under study (Table 1). The predominant pattern - 11 of 21 isolates - was characterized by resistance to ampicillin, streptomycin, tetracycline and nalidixic acid, with the addition of kanamycin in two cases. Since 2001 the ampicillin-sulfonamides resistance pattern became more frequent. Two strains only were susceptible to all the antimicrobial agents tested.

Six plasmid profiles - a to f - were identified among 19 of the 21 isolates (Table 1 and Fig. 1). The two fully susceptible isolates were also plasmid free.
Three distinct XbaI PFGE patterns, designated X-A to X-C were observed (Fig. 2). Within PFGE type X-A pattern there were three subtypes X-A1, X-A2 and X-A3. PFGE subtype X-A1 accounted for 16 of 21 isolates, whilst subtypes X-A2 and X-A3 were represented by a single isolate, respectively (Table 1). The DNAs of the 16 strains with XbaI-PFGE pattern X-A1 were digested by BlnI. Fourteen were indistinguishable and were assigned the pattern B-A1; the remaining two were assigned two different subtypes B-A2 and B-A3, differing from B-A1 by two bands each (data not shown).

Discussion
The strains of S. Heidelberg investigated in this study cannot be considered as a representative sample of the strains circulating in Italy because of the selection method used. This is an inherent limit of passive surveillance systems, such as those set up in most European countries for enteric pathogens, that depend on voluntary adherence by peripheral laboratories. Nevertheless, these surveillance networks are often able to trace the transmission routes of some Salmonella clones and describe time and space trends of the serotypes of major interest to Public Health.
PFGE patterns appear to be consistent with the dissemination of a common outbreak strain, though different drug susceptibility and plasmid profiles have been found. PFGE analysis has indeed proved to be a very effective tool in determining whether some isolates are essentially clonal, and its application as the basic method in molecular surveillance networks of verocytotoxigenic Escherichia coli and Salmonella has already been successful for some years (11). Furthermore, the results of a recent study on epidemiology of S. Heidelberg in the US have demonstrated the presence of a clonal strain over a ten year period, thus suggesting the possible persistence of a particular strain over a wide area for a prolonged period (12). Moreover, the results obtained concur with reports by other authors who have used PFGE to identify clonal relationship between Salmonella isolates, but found different antimicrobial resistance patterns and plasmid profiles within an apparently unique chromosomal clone (6,13,14). In our case too, the intrinsic instability of extra-chromosomal DNA, the selective pressure by use of different antibiotics in different places, and the broad period of observation might justify the heterogeneity of the Heidelberg isolates on the basis of their plasmid and drug resistance patterns. On the other hand, more strain-specific markers, such as the drug resistance pattern/plasmid profile, might be more successful than PFGE alone in delineating local transmission routes triggered by clonally spreading Salmonella serotypes, such as Enteritidis and, presumably, Heidelberg.

The association of such a serotype with avian hosts and eggs has been known for some time. S. Heidelberg was the predominant serotype recovered from ovary samples in a survey of layer flocks in 1991, and its ability to penetrate and grow into the interior of hens' eggs has been well documented (3,15). Moreover, the high prevalence of nalidixic acid resistance within strains of S. Heidelberg, a zoonotic serotype closely associated with chickens and turkeys, could be related to their ecological niche. Indeed, the use of fluoroquinolones is common in the poultry industry and is temporally strongly associated with raising frequencies of resistances in many Salmonella serotypes, including Enteritidis (16,17).
Clonal diffusion of some predominant genotypes may be more widespread than is currently recognized and involve Salmonella strains other than the traditional Enteritidis and Typhimurium DT104 complex. In Italy, Heidelberg serotype is apparently able to spread clonally nationwide through the poultry vehicle. A larger study involving a more representative sample of strains from Italy and, possibly, other European countries might more confidently evaluate epidemiological features of this Salmonella serotype. Molecular epidemiological monitoring should be a routine tool for detection, and quantitative assessment of unexpected events related to zoonotic serotypes.